Red Hat Enterprise Linux 6

Transcription

1 Red Hat Enterprise Linux 6 Virtualization Tuning and Optimization Guide Optimizing your virtual environment Last Updated:

2

3 Red Hat Enterprise Linux 6 Virtualization Tuning and Optimization Guide Optimizing your virtual environment Jiri Herrmann Red Hat Customer Content Services jherrman@redhat.com Yehuda Zimmerman Red Hat Customer Content Services yzimmerm@redhat.com Dayle Parker Red Hat Customer Content Services Scott Radvan Red Hat Customer Content Services Red Hat Subject Matter Experts

4 Legal Notice Copyright 2017 Red Hat, Inc. This document is licensed by Red Hat under the Creative Commons Attribution-ShareAlike 3.0 Unported License. If you distribute this document, or a modified version of it, you must provide attribution to Red Hat, Inc. and provide a link to the original. If the document is modified, all Red Hat trademarks must be removed. Red Hat, as the licensor of this document, waives the right to enforce, and agrees not to assert, Section 4d of CC-BY-SA to the fullest extent permitted by applicable law. Red Hat, Red Hat Enterprise Linux, the Shadowman logo, JBoss, OpenShift, Fedora, the Infinity logo, and RHCE are trademarks of Red Hat, Inc., registered in the United States and other countries. Linux is the registered trademark of Linus Torvalds in the United States and other countries. Java is a registered trademark of Oracle and/or its affiliates. XFS is a trademark of Silicon Graphics International Corp. or its subsidiaries in the United States and/or other countries. MySQL is a registered trademark of MySQL AB in the United States, the European Union and other countries. Node.js is an official trademark of Joyent. Red Hat Software Collections is not formally related to or endorsed by the official Joyent Node.js open source or commercial project. The OpenStack Word Mark and OpenStack logo are either registered trademarks/service marks or trademarks/service marks of the OpenStack Foundation, in the United States and other countries and are used with the OpenStack Foundation's permission. We are not affiliated with, endorsed or sponsored by the OpenStack Foundation, or the OpenStack community. All other trademarks are the property of their respective owners. Abstract The Red Hat Enterprise Linux Virtualization Tuning and Optimization Guide covers KVM and virtualization performance. Within this guide you can find tips and suggestions for making full use of KVM performance features and options for your host systems and guest virtual machines.

5 Table of Contents Table of Contents. CHAPTER INTRODUCTION ABOUT THIS GUIDE KVM PERFORMANCE ARCHITECTURE OVERVIEW PERFORMANCE FEATURES AND IMPROVEMENTS 4. CHAPTER PERFORMANCE MONITORING TOOLS INTRODUCTION PERF KVM 6. CHAPTER OPTIMIZING VIRTUALIZATION PERFORMANCE WITH..... VIRT-MANAGER INTRODUCTION OPERATING SYSTEM DETAILS AND DEVICES Specifying Guest Virtual Machine Details Remove Unused Devices CPU PERFORMANCE OPTIONS Option: Available CPUs Option: CPU Configuration Option: CPU Topology Option: CPU Pinning VIRTUAL DISK PERFORMANCE OPTIONS 14. CHAPTER TUNED INTRODUCTION TUNED AND TUNED-ADM 15. CHAPTER NETWORKING INTRODUCTION NETWORK TUNING TIPS VIRTIO AND VHOST_NET DEVICE ASSIGNMENT AND SR-IOV 18. CHAPTER I/O... SCHEDULING I/O SCHEDULING I/O SCHEDULING WITH RED HAT ENTERPRISE LINUX AS A VIRTUALIZATION HOST I/O SCHEDULING WITH RED HAT ENTERPRISE LINUX AS A VIRTUALIZATION GUEST CONFIGURING THE I/O SCHEDULER Configuring the I/O Scheduler for Red Hat Enterprise Linux 5 and Configuring the I/O Scheduler for Red Hat Enterprise Linux CHAPTER BLOCK I/O INTRODUCTION BLOCK I/O TUNING CACHING I/O MODE 22. CHAPTER MEMORY INTRODUCTION MEMORY TUNING TIPS MEMORY TUNING ON VIRTUAL MACHINES Memory Monitoring Tools Memory Tuning with virsh memtune Memory Backing 26 1

6 Virtualization Tuning and Optimization Guide Huge Pages and Transparent Huge Pages (THP) Configuring Transparent Huge Pages Configuring Static Huge Pages Enabling 1 GB huge pages for guests at boot or runtime 8.4. KERNEL SAME-PAGE MERGING (KSM) The KSM Service The KSM Tuning Service KSM Variables and Monitoring Deactivating KSM CHAPTER NUMA INTRODUCTION NUMA MEMORY ALLOCATION POLICIES LIBVIRT NUMA TUNING NUMA vcpu Pinning Domain Processes Domain vcpu Threads Using emulatorpin Tuning vcpu Pinning with virsh Tuning Domain Process CPU Pinning with virsh Tuning Domain Process Memory Policy with virsh NUMA-AWARE KERNEL SAMEPAGE MERGING (KSM) 40. APPENDIX A.... REVISION HISTORY

7 Table of Contents 3

8 Virtualization Tuning and Optimization Guide CHAPTER 1. INTRODUCTION 1.1. ABOUT THIS GUIDE The Red Hat Enterprise Linux Virtualization Tuning and Optimization Guide contains details of configurable options and settings, and other suggestions that will help you achieve optimal performance of your Red Hat Enterprise Linux hosts and guest virtual machines. Following this introduction, the guide consists of the following sections: Virt-manager tuned Networking Memory Block I/O NUMA Performance Monitoring Tools 1.2. KVM PERFORMANCE ARCHITECTURE OVERVIEW The following points provide a brief overview of KVM as it pertains to system performance and process/thread management: When using KVM, guests run as a Linux process on the host. Virtual CPUs (vcpus) are implemented as normal threads, handled by the Linux scheduler. Guests inherit features such as NUMA and Huge Pages from the kernel. Disk and network I/O settings in the host have a significant performance impact. Network traffic typically travels through a software-based bridge PERFORMANCE FEATURES AND IMPROVEMENTS CPU/Kernel Memory NUMA - Non-Uniform Memory Access. See Chapter 9, NUMA for details on NUMA. CFS - Completely Fair Scheduler. A modern class-focused scheduler. RCU - Read Copy Update. Better handling of shared thread data. Up to 160 virtual CPUs (vcpus). Huge Pages and other optimizations for memory-intensive environments. See Chapter 8, Memory for details. 4

9 CHAPTER 1. INTRODUCTION Networking vhost-net - a fast, kernel-based VirtIO solution. SR-IOV - for near-native networking performance levels. Block I/O AIO - Support for a thread to overlap other I/O operations. MSI - PCI bus device interrupt generation. Scatter Gather - An improved I/O mode for data buffer handling. NOTE For more details on virtualization support, limits, and features, refer to the Red Hat Enterprise Linux 6 Virtualization Getting Started Guide and the following URLs: 5

10 Virtualization Tuning and Optimization Guide CHAPTER 2. PERFORMANCE MONITORING TOOLS 2.1. INTRODUCTION This chapter describes tools used to monitor guest virtual machine environments PERF KVM You can use the perf command with the kvm option to collect guest operating system statistics from the host. In Red Hat Enterprise Linux, the perf package provides the perf command. Run rpm -q perf to see if the perf package is installed. If it is not installed, and you want to install it to collect and analyze guest operating system statistics, run the following command as the root user: yum install perf In order to use perf kvm in the host, you must have access to the /proc/modules and /proc/kallsyms files from the guest. There are two methods to achieve this. Refer to the following procedure, Procedure 2.1, Copying /proc files from guest to host to transfer the files into the host and run reports on the files. Alternatively, refer to Procedure 2.2, Alternative: using sshfs to directly access files to directly mount the guest and access the files. Procedure 2.1. Copying /proc files from guest to host IMPORTANT If you directly copy the required files (for instance, via scp) you will only copy files of zero length. This procedure describes how to first save the files in the guest to a temporary location (with the cat command), and then copy them to the host for use by perf kvm. 1. Log in to the guest and save files Log in to the guest and save /proc/modules and /proc/kallsyms to a temporary location, /tmp: # cat /proc/modules > /tmp/modules # cat /proc/kallsyms > /tmp/kallsyms 2. Copy the temporary files to the host Once you have logged off from the guest, run the following example scp commands to copy the saved files to the host. You should substitute your host name and TCP port if they are different: # scp root@guestmachine:/tmp/kallsyms guest-kallsyms # scp root@guestmachine:/tmp/modules guest-modules You now have two files from the guest (guest-kallsyms and guest-modules) on the host, ready for use by perf kvm. 3. Recording and reporting events with perf kvm Using the files obtained in the previous steps, recording and reporting of events in the guest, the host, or both is now possible. 6

11 CHAPTER 2. PERFORMANCE MONITORING TOOLS Run the following example command: # perf kvm --host --guest --guestkallsyms=guest-kallsyms \ --guestmodules=guest-modules record -a -o perf.data NOTE If both --host and --guest are used in the command, output will be stored in perf.data.kvm. If only --host is used, the file will be named perf.data.host. Similarly, if only --guest is used, the file will be named perf.data.guest. Pressing Ctrl-C stops recording. 4. Reporting events The following example command uses the file obtained by the recording process, and redirects the output into a new file, analyze. perf kvm --host --guest --guestmodules=guest-modules report -i perf.data.kvm \ --force > analyze View the contents of the analyze file to examine the recorded events: # cat analyze # Events: 7K cycles # # Overhead Command Shared Object Symbol # # 95.06% vi vi [.] 0x % init [kernel.kallsyms] [k] intel_idle 0.36% vi libc-2.12.so [.] _wordcopy_fwd_aligned 0.32% vi libc-2.12.so [.] strlen_sse % swapper [kernel.kallsyms] [k] intel_idle 0.13% init [kernel.kallsyms] [k] uhci_irq 0.11% perf [kernel.kallsyms] [k] generic_exec_single 0.11% init [kernel.kallsyms] [k] tg_shares_up 0.10% qemu-kvm [kernel.kallsyms] [k] tg_shares_up [output truncated...] Procedure 2.2. Alternative: using sshfs to directly access files 7

12 Virtualization Tuning and Optimization Guide IMPORTANT This is provided as an example only. You will need to substitute values according to your environment. # Get the PID of the qemu process for the guest: PID=`ps -eo pid,cmd grep "qemu.*-name GuestMachine" \ grep -v grep awk '{print $1}'` # Create mount point and mount guest mkdir -p /tmp/guestmount/$pid sshfs -o allow_other,direct_io GuestMachine:/ /tmp/guestmount/$pid # Begin recording perf kvm --host --guest --guestmount=/tmp/guestmount \ record -a -o perf.data # Ctrl-C interrupts recording. Run report: perf kvm --host --guest --guestmount=/tmp/guestmount report \ -i perf.data # Unmount sshfs to the guest once finished: fusermount -u /tmp/guestmount 8

13 CHAPTER 3. OPTIMIZING VIRTUALIZATION PERFORMANCE WITH VIRT-MANAGER CHAPTER 3. OPTIMIZING VIRTUALIZATION PERFORMANCE WITH VIRT-MANAGER 3.1. INTRODUCTION This chapter covers performance tuning options for virt-manager, a desktop tool for managing guest virtual machines OPERATING SYSTEM DETAILS AND DEVICES Specifying Guest Virtual Machine Details The virt-manager tool provides different profiles depending on what operating system type and version are selected for a new guest virtual machine. When creating a guest, you should provide as many details as possible; this can improve performance by enabling features available for your specific type of guest. Refer to the following example screen capture of the virt-manager tool. When creating a new guest virtual machine, always specify your intended OS type and Version: Figure 3.1. Provide the OS type and Version Remove Unused Devices Removing unused or unnecessary devices can improve performance. For instance, a guest tasked as a web server is unlikely to require audio features or an attached tablet. 9

14 Virtualization Tuning and Optimization Guide Refer to the following example screen capture of the virt-manager tool. Click the Remove button to remove unnecessary devices: Figure 3.2. Remove unused devices 3.3. CPU PERFORMANCE OPTIONS Several CPU related options are available to your guest virtual machines. Configured correctly, these options can have a large impact on performance. The following image shows the CPU options available to your guests. The remainder of this section shows and explains the impact of these options. 10

15 CHAPTER 3. OPTIMIZING VIRTUALIZATION PERFORMANCE WITH VIRT-MANAGER Figure 3.3. CPU Performance Options Option: Available CPUs Use this option to adjust the amount of virtual CPUs (vcpus) available to the guest. It is also possible to change the number of vcpus while the guest is running, which is referred to as hot plugging and hot unplugging. IMPORTANT The hot unplugging feature is only available as a Technology Preview. Therefore, it is not supported and not recommended for use in high-value deployments. If you allocate more than is available on the host (known as overcommitting), a warning is displayed, as shown in the following image: 11

16 Virtualization Tuning and Optimization Guide Figure 3.4. CPU overcommit CPUs are overcommitted when the sum of vcpus for all guests on the system is greater than the number of host CPUs on the system. You can overcommit CPUs with one or multiple guests if the total number of vcpus is greater than the number of host CPUs. IMPORTANT As with memory overcommitting, CPU overcommitting can have a negative impact on performance, for example in situations with a heavy or unpredictable guest workload. Refer to the Red Hat Enterprise Linux 6 Virtualization Administration Guide for more details on overcommitting Option: CPU Configuration Use this option to select the CPU configuration type, based on the desired CPU model. Expand the list to see available options, or click the Copy host CPU configuration button to detect and apply the physical host's CPU model and configuration. Once you select a CPU configuration, its available CPU features/instructions are displayed and can be individually enabled/disabled in the CPU Features list. Refer to the following diagram which shows these options: Figure 3.5. CPU Configuration Options NOTE Copying the host CPU configuration is recommended over manual configuration. 12

17 CHAPTER 3. OPTIMIZING VIRTUALIZATION PERFORMANCE WITH VIRT-MANAGER NOTE Alternately, run the virsh capabilities command on your host machine to view the virtualization capabilities of your system, including CPU types and NUMA capabilities Option: CPU Topology Use this option to apply a particular CPU topology (Sockets, Cores, Threads) to the virtual CPUs for your guest virtual machine. Refer to the following diagram which shows an example of this option: Figure 3.6. CPU Topology Options NOTE Although your environment may dictate other requirements, selecting any desired number of sockets, but with only a single core and a single thread usually gives the best performance results Option: CPU Pinning Large performance improvements can be obtained by adhering to the system's specific NUMA topology. Use this option to automatically generate a pinning configuration that is valid for the host. Figure 3.7. CPU Pinning 13

18 Virtualization Tuning and Optimization Guide WARNING Do not use this option if the guest has more vcpus than a single NUMA node. Using the Pinning option will constrain the guest's vcpu threads to a single NUMA node; however, threads will be able to move around within that NUMA node. For tighter binding capabilities, use the output from the lscpu command to establish a 1:1 physical CPU to vcpu binding using virsh cpupin. Refer to Chapter 9, NUMA for more information on NUMA and CPU pinning VIRTUAL DISK PERFORMANCE OPTIONS Several virtual disk related options are available to your guest virtual machines during installation that can impact performance. The following image shows the virtual disk options available to your guests. The cache mode, IO mode, and IO tuning can be selected in the Virtual Disk section in virtmanager. Set these parameters in the fields under Performance options, as shown in the following image: Figure 3.8. Virtual Disk Performance Options IMPORTANT When setting the virtual disk performance options in virt-manager, the virtual machine must be restarted for the settings to take effect. See Section 7.3, Caching and Section 7.4, I/O Mode for descriptions of these settings and instructions for editing these settings in the guest XML configuration. 14

19 CHAPTER 4. TUNED CHAPTER 4. TUNED 4.1. INTRODUCTION This chapter covers using tuned daemon for dynamically tuning system settings in virtualized environments TUNED AND TUNED-ADM tuned is a tuning profile delivery mechanism that adapts Red Hat Enterprise Linux for certain workload characteristics, such as requirements for CPU-intensive tasks, or storage/network throughput responsiveness. It can react to changes in CPU and network use and adjust settings to improve performance in active devices, or reduce power consumption in inactive devices. The accompanying ktune partners with the tuned-adm tool to provide a number of tuning profiles that are pre-configured to enhance performance and reduce power consumption in a number of specific use cases. Edit these profiles or create new profiles to create performance solutions tailored to your environment. The virtualization-related profiles provided as part of tuned-adm include: virtual-guest Based on the enterprise-storage profile, virtual-guest also decreases the swappiness of virtual memory. This profile is available in Red Hat Enterprise Linux 6.3 and later, and is the recommended profile for guest machines. virtual-host Based on the enterprise-storage profile, virtual-host also decreases the swappiness of virtual memory and enables more aggressive writeback of dirty pages. This profile is available in Red Hat Enterprise Linux 6.3 and later, and is the recommended profile for virtualization hosts, including both KVM and Red Hat Enterprise Virtualization hosts. Install the tuned package and its associated systemtap scripts with the command: yum install tuned Installing the tuned package also sets up a sample configuration file at /etc/tuned.conf and activates the default profile. Start tuned by running: service tuned start To start tuned every time the machine boots, run: chkconfig tuned on To list all available profiles and identify the current active profile, run: # tuned-adm list Available profiles: 15

20 Virtualization Tuning and Optimization Guide - latency-performance - sap - virtual-guest - laptop-battery-powersave - virtual-host - enterprise-storage - desktop-powersave - spindown-disk - default - throughput-performance - laptop-ac-powersave - server-powersave Current active profile: default To only display the currently active profile, run: tuned-adm active To switch to one of the available profiles, run: tuned-adm profile profile_name For example, to switch to the virtual-host profile, run: tuned-adm profile virtual-host NOTE It is possible to create custom tuned profiles to encapsulate a set of tuning parameters. For instructions on creating custom tuned profiles, refer to the tuned.conf man page. In some cases, it is preferable to disable tuned to use parameters set manually. To disable all tuning, run: tuned-adm off NOTE Refer to the Red Hat Enterprise Linux 6 Power Management Guide for further information about tuned, tuned-adm and ktune. 16

21 CHAPTER 5. NETWORKING CHAPTER 5. NETWORKING 5.1. INTRODUCTION This chapter covers network optimization topics for virtualized environments NETWORK TUNING TIPS Use multiple networks to avoid congestion on a single network. For example, have dedicated networks for management, backups and/or live migration. Usually, matching the default MTU (1500 bytes) in all components is sufficient. If you require larger messages, increasing the MTU value can reduce fragmentation. If you change the MTU, all devices in the path should have a matching MTU value. Use arp_filter to prevent ARP Flux, an undesirable condition that can occur in both hosts and guests and is caused by the machine responding to ARP requests from more than one network interface: echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter or edit /etc/sysctl.conf to make this setting persistent. NOTE Refer to the following URL for more information on ARP Flux: VIRTIO AND VHOST_NET The following diagram demonstrates the involvement of the kernel in the Virtio and vhost_net architectures. Figure 5.1. Virtio and vhost_net architectures 17

22 Virtualization Tuning and Optimization Guide vhost_net moves part of the Virtio driver from the user space into the kernel. This reduces copy operations, lowers latency and CPU usage DEVICE ASSIGNMENT AND SR-IOV The following diagram demonstrates the involvement of the kernel in the Device Assignment and SR- IOV architectures. Figure 5.2. Device assignment and SR-IOV Device assignment presents the entire device to the guest. SR-IOV needs support in drivers and hardware, including the NIC and the system board and allows multiple virtual devices to be created and passed into different guests. A vendor-specific driver is required in the guest, however, SR-IOV offers the lowest latency of any network option. 18

23 CHAPTER 6. I/O SCHEDULING CHAPTER 6. I/O SCHEDULING This chapter covers disk performance optimization topics for virtualized environments I/O SCHEDULING You can use an I/O scheduler to improve disk performance both when Red Hat Enterprise Linux is the virtualization host and when it is the virtualization guest I/O SCHEDULING WITH RED HAT ENTERPRISE LINUX AS A VIRTUALIZATION HOST When using Red Hat Enterprise Linux as a host for virtualized guests, the default cfq scheduler is usually ideal. This scheduler performs well on nearly all workloads. However, if minimizing I/O latency is more important than maximizing I/O throughput on the guest workloads, it may be beneficial to use the deadline scheduler. The deadline scheduler is also the scheduler used by the tuned profile virtual-host I/O SCHEDULING WITH RED HAT ENTERPRISE LINUX AS A VIRTUALIZATION GUEST You can use I/O scheduling on a Red Hat Enterprise Linux guest virtual machine, regardless of the hypervisor on which the guest is running. The following is a list of benefits and issues that should be considered: Red Hat Enterprise Linux guests often benefit greatly from the noop scheduler. This noop scheduler allows the host machine or hypervisor to optimize the input/output requests. The noop scheduler can combine small requests from the guest operating system into larger requests, before handing the I/O to the hypervisor. However, noop tries to use the fewest number of CPU cycles in the guest for I/O scheduling. The host/hypervisor has an overview of the requests of all guests and uses a separate strategy for handling I/O. NOTE For Red Hat Enterprise Linux 7.2 and newer, virtio-blk always uses noop implicitly. This is because it uses blk-mq. Depending on the workload I/O and how storage devices are attached, schedulers like deadline can be more advantageous. Performance testing is required to verify which scheduler is the best one to use in each situation. All scheduler tuning should be tested under normal operating conditions, as synthetic benchmarks typically do not accurately compare performance of systems using shared resources in virtual environments. Guests using storage accessed by iscsi, SR-IOV, or physical device passthrough should not use the noop scheduler. These methods do not allow the host to optimize I/O requests to the underlying physical device. In virtualized environments, sometimes it is not beneficial to schedule I/O on both the host and guest layers. If multiple guests use storage on a filesystem or block device managed by the host operating 19

24 Virtualization Tuning and Optimization Guide system, the host may be able to schedule I/O more efficiently because it is aware of requests from all guests and knows the physical layout of storage, which may not map linearly to the guests' virtual storage. On the other hand, depending on the workload, it may be beneficial to use a scheduler like deadline in the guest virtual machine CONFIGURING THE I/O SCHEDULER The I/O scheduler is configured one way for Red Hat Enterprise Linux 5 and 6 and a different way for Red Hat Enterprise Linux Configuring the I/O Scheduler for Red Hat Enterprise Linux 5 and 6 The I/O scheduler can be selected at boot time using the elevator kernel parameter. In the following example grub.conf stanza, the system is configured to use the noop scheduler. The example shown is for VMware ESX. title Red Hat Enterprise Linux Server ( el5) root (hd0,0) kernel /vmlinuz el5 ro root=/dev/vg0/lv0 elevator=noop initrd /initrd el5.img Configuring the I/O Scheduler for Red Hat Enterprise Linux 7 For a Red Hat Enterprise Linux 7 guest, changes to the I/O scheduler can be made persistent through boot by adding the following elevator to /etc/grub2.cfg, as in the following example: # vi /etc/grub2.cfg linux16 /vmlinuz-kernel-version root=/dev/mapper/vg0-lv0 ro rd.lvm.lv=vg0/lv0 vconsole.keymap=us vconsole.font=latarcyrheb-sun16 rhgb quiet elevator=deadline initrd16 /initramfs-kernel-version.img 20

25 CHAPTER 7. BLOCK I/O CHAPTER 7. BLOCK I/O 7.1. INTRODUCTION This chapter covers optimizing I/O settings in virtualized environments BLOCK I/O TUNING The virsh blkiotune command allows administrators to set or display a guest virtual machine's block I/O parameters manually in the <blkio> element in the guest XML configuration. To display current <blkio> parameters for a virtual machine: # virsh blkiotune virtual_machine To set a virtual machine's <blkio> parameters, refer to the following command and replace values according to your environment: # virsh blkiotune virtual_machine [--weight number] [--device-weights string] [--config] [--live] [--current] Parameters include: weight The I/O weight, within the range 100 to device-weights A single string listing one or more device/weight pairs, in the format of /path/to/device,weight,/path/to/device,weight. Each weight must be within the range , or the value 0 to remove that device from per-device listings. Only the devices listed in the string are modified; any existing per-device weights for other devices remain unchanged. config Add the --config option for changes to take effect at next boot. live Add the --live option to apply the changes to the running virtual machine. NOTE The --live option requires the hypervisor to support this action. Not all hypervisors allow live changes of the maximum memory limit. current Add the --current option to apply the changes to the current virtual machine. 21

26 Virtualization Tuning and Optimization Guide NOTE See # virsh help blkiotune for more information on using the virsh blkiotune command CACHING Caching options can be configured with virt-manager during guest installation, or on an existing guest virtual machine by editing the guest XML configuration. Table 7.1. Caching options Caching Option Description Cache=none I/O from the guest is not cached on the host, but may be kept in a writeback disk cache. Use this option for guests with large I/O requirements. This option is generally the best choice, and is the only option to support migration. Cache=writethrough I/O from the guest is cached on the host but written through to the physical medium. This mode is slower and prone to scaling problems. Best used for small number of guests with lower I/O requirements. Suggested for guests that do not support a writeback cache (such as Red Hat Enterprise Linux 5.5 and earlier), where migration is not needed. Cache=writeback I/O from the guest is cached on the host. To configure the cache mode in the guest XML, use virsh edit to edit the cache setting inside the driver tag, specifying none, writeback, or writethrough. For example, to set the cache as writeback: <disk type='file' device='disk'> <driver name='qemu' type='raw' cache='writeback'/> 7.4. I/O MODE I/O mode options can be configured on a virtual machine during installation with virt-manager or the virt-install command, or on an existing guest by editing the guest XML configuration. Table 7.2. IO mode options Caching Option Description IO=native The default for Red Hat Enterprise Virtualization environments. This mode uses kernel asynchronous I/O with direct I/O options. 22

27 CHAPTER 7. BLOCK I/O Caching Option Description IO=threads Sets the I/O mode to host user-mode based threads. IO=default Sets the I/O mode to the kernel default. In Red Hat Enterprise Linux 6, the default is IO=threads. In virt-manager, the I/O mode can be specified under Virtual Disk. For information on using virtmanager to change the I/O mode, see Section 3.4, Virtual Disk Performance Options To configure the I/O mode in the guest XML, use virsh edit to edit the io setting inside the driver tag, specifying native, threads, or default. For example, to set the I/O mode to threads: <disk type='file' device='disk'> <driver name='qemu' type='raw' io='threads'/> To configure the I/O mode when installing a guest using virt-install, add the io option to the -- disk path parameter. For example, to configure io=threads during guest installation: virt-install \ --disk path=/storage/images/$name.img,io=threads,opt2=val

28 Virtualization Tuning and Optimization Guide CHAPTER 8. MEMORY 8.1. INTRODUCTION This chapter covers memory optimization options for virtualized environments MEMORY TUNING TIPS To optimize memory performance in a virtualized environment, consider the following: Do not allocate more resources to guest than it will use. If possible, assign a guest to a single NUMA node, providing that resources are sufficient on that NUMA node. For more information on using NUMA, see Chapter 9, NUMA. When increasing the amount of memory a guest virtual machine can use while the guest is running, also referred to as hot plugging, the memory needs to be manually brought online on the guest by one of the following methods: Create a custom udev rule 1. Create a file with a name that ends in the.rulessuffix in the /etc/udev/rules.d/ directory: # touch /etc/udev/rules.d/rulename.rules 2. Add the memory onlining to the created file: # echo 'SUBSYSTEM=="memory", ACTION=="add", ATTR{state}=="offline", ATTR{state}="online"' > /etc/udev/rules.d/rulename.rules 3. Reload udev rules: # udevadm control --reload Bring inactive memory online manually after each hot plug # for mblock in /sys/devices/system/memory/memory*; do echo online > $mblock/state; done 8.3. MEMORY TUNING ON VIRTUAL MACHINES Memory Monitoring Tools Memory usage can be monitored in virtual machines using tools used in bare metal environments. Tools useful for monitoring memory usage and diagnosing memory-related problems include: top vmstat 24

29 CHAPTER 8. MEMORY numastat /proc/ NOTE For details on using these performance tools, refer to the Red Hat Enterprise Linux 6 Performance Tuning Guide and the man pages for these commands Memory Tuning with virsh memtune The optional <memtune> element in the guest XML configuration allows administrators to configure guest virtual machine memory settings manually. If <memtune> is omitted, default memory settings apply. Display or set memory parameters in the <memtune> element in a virtual machine with the virsh memtune command, replacing values according to your environment: virsh # memtune virtual_machine --parameter size Available parameters include: hard-limit The maximum memory the virtual machine can use, in kibibytes (blocks of 1024 bytes). WARNING Setting this limit too low can result in the virtual machine being killed by the kernel. soft-limit The memory limit to enforce during memory contention, in kibibytes (blocks of 1024 bytes). swap-hard-limit The maximum memory plus swap the virtual machine can use, in kibibytes (blocks of 1024 bytes). The swap-hard-limit value must be more than the hard-limit value. min-guarantee The guaranteed minimum memory allocation for the virtual machine, in kibibytes (blocks of 1024 bytes). 25

30 Virtualization Tuning and Optimization Guide NOTE See # virsh help memtune for more information on using the virsh # memtune command Memory Backing The optional <memorybacking> element may contain several elements that influence how virtual memory pages are backed by host pages. Red Hat Enterprise Linux 6.6 and later includes support for locking a virtual machine's memory pages in the host memory with the locked parameter. Setting locked prevents the host from swapping out memory pages belonging to the guest, which can cause latency. Add the following to the guest XML using the virsh edit command to lock the virtual memory pages in the host's memory: <memorybacking> <locked/> </memorybacking> IMPORTANT When setting locked, a hard_limit must be set in the <memtune> element to the maximum memory configured for the guest, plus any memory consumed by the process itself. Setting nosharepages prevents the host from merging the same memory used among guests. To instruct the hypervisor to disable share pages for a guest, add the following to the guest's XML using the virsh edit command: <memorybacking> <nosharepages/> </memorybacking> Huge Pages and Transparent Huge Pages (THP) x86 CPUs usually address memory in 4kB pages, but they are capable of using larger 2 MB or 1 GB pages known as huge pages. KVM guests can be deployed with huge page memory support in order to improve performance by increasing CPU cache hits against the Transaction Lookaside Buffer (TLB). A kernel feature enabled by default in Red Hat Enterprise Linux 6, huge pages can significantly increase performance, particularly for large memory and memory-intensive workloads. Red Hat Enterprise Linux 6 is able to more effectively manage large amounts of memory by increasing the page size through the use of huge pages. Red Hat Enterprise Linux 6.7 systems support both 2 MB and 1 GB huge pages, which can be allocated at boot or at runtime. See Section , Enabling 1 GB huge pages for guests at boot or runtime for instructions on enabling multiple huge page sizes Configuring Transparent Huge Pages Transparent huge pages (THP) automatically optimize system settings for performance. By allowing all 26

31 CHAPTER 8. MEMORY free memory to be used as cache, performance is increased. As KSM can reduce the occurence of transparent huge pages, you may want to disable it before enabling THP. If you want to disable KSM, refer to Section 8.4.4, Deactivating KSM. Transparent huge pages are enabled by default. To check the current status, run: # cat /sys/kernel/mm/transparent_hugepage/enabled To enable transparent huge pages to be used by default, run: # echo always > /sys/kernel/mm/transparent_hugepage/enabled This will set /sys/kernel/mm/transparent_hugepage/enabled to always. To disable transparent huge pages: # echo never > /sys/kernel/mm/transparent_hugepage/enabled Transparent Huge Page support does not prevent the use of static huge pages. However, when static huge pages are not used, KVM will use transparent huge pages instead of the regular 4kB page size Configuring Static Huge Pages In some cases, greater control of huge pages is preferable. To use static huge pages on guests, add the following to the guest XML configuration: <memorybacking> <hugepages/> </memorybacking> This instructs the host to allocate memory to the guest using huge pages, instead of using the default page size. To view the current huge pages value, run the following command: cat /proc/sys/vm/nr_hugepages Procedure 8.1. Setting huge pages The following example procedure shows the commands to set huge pages. 1. View the current huge pages value: # cat /proc/meminfo grep Huge AnonHugePages: 2048 kb HugePages_Total: 0 HugePages_Free: 0 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 2048 kb 2. Huge pages are set in increments of 2MB. To set the number of huge pages to 25000, use the following command: 27

32 Virtualization Tuning and Optimization Guide echo > /proc/sys/vm/nr_hugepages NOTE To make the setting persistent, add the following lines to the /etc/sysctl.conf file on the guest machine, with X being the intended number of huge pages: # echo 'vm.nr_hugepages = X' >> /etc/sysctl.conf # sysctl -p Afterwards, add transparent_hugepage=never to the kernel boot parameters by appending it to the end of the /kernel line in the /etc/grub2.cfg file on the guest. 3. Mount the huge pages: # mount -t hugetlbfs hugetlbfs /dev/hugepages 4. Restart libvirtd, then restart the virtual machine: # service libvirtd restart Stopping libvirtd daemon: [ OK ] Starting libvirtd daemon: [ OK ] # virsh start virtual_machine 5. Verify the changes in /proc/meminfo: # cat /proc/meminfo grep Huge AnonHugePages: 0 kb HugePages_Total: HugePages_Free: HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 2048 kb Huge pages can benefit not only the host but also guests, however, their total huge pages value must be less than what is available in the host Enabling 1 GB huge pages for guests at boot or runtime Red Hat Enterprise Linux 6.7 systems support 2 MB and 1 GB huge pages, which can be allocated at boot or at runtime. Procedure 8.2. Allocating 1 GB huge pages at boot time 1. To allocate different sizes of huge pages at boot, use the following command, specifying the number of huge pages. This example allocates 4 1 GB huge pages and MB huge pages: 28

33 CHAPTER 8. MEMORY 'default_hugepagesz=1g hugepagesz=1g hugepages=4 hugepagesz=2m hugepages=1024' Change this command line to specify a different number of huge pages to be allocated at boot. NOTE The next two steps must also be completed the first time you allocate 1 GB huge pages at boot time. 2. Mount the 2 MB and 1 GB huge pages on the host: # mkdir /dev/hugepages1g # mount -t hugetlbfs -o pagesize=1g none /dev/hugepages1g # mkdir /dev/hugepages2m # mount -t hugetlbfs -o pagesize=2m none /dev/hugepages2m 3. Restart libvirtd to enable the use of 1 GB huge pages on guests: # service restart libvirtd 1 GB huge pages are now available for guests. Procedure 8.3. Allocating 1 GB huge pages at runtime 1 GB huge pages can also be allocated at runtime. Runtime allocation allows the system administrator to choose which NUMA node to allocate those pages from. However, runtime page allocation can be more prone to allocation failure than boot time allocation due to memory fragmentation. 1. To allocate different sizes of huge pages at runtime, use the following command, replacing values for the number of huge pages, the NUMA node to allocate them from, and the huge page size: # echo 4 > /sys/devices/system/node/node1/hugepages/hugepages kB/nr_hugepages # echo 1024 > /sys/devices/system/node/node3/hugepages/hugepages- 2048kB/nr_hugepages This example command allocates 4 1 GB huge pages from node1 and MB huge pages from node3. These huge page settings can be changed at any time with the above command, depending on the amount of free memory on the host system. NOTE The next two steps must also be completed the first time you allocate 1 GB huge pages at runtime. 2. Mount the 2 MB and 1 GB huge pages on the host: # mkdir /dev/hugepages1g 29

34 Virtualization Tuning and Optimization Guide # mount -t hugetlbfs -o pagesize=1g none /dev/hugepages1g # mkdir /dev/hugepages2m # mount -t hugetlbfs -o pagesize=2m none /dev/hugepages2m 3. Restart libvirtd to enable the use of 1 GB huge pages on guests: # systemctl restart libvirtd 8.4. KERNEL SAME-PAGE MERGING (KSM) Kernel same-page Merging (KSM), used by the KVM hypervisor, allows KVM guests to share identical memory pages. These shared pages are usually common libraries or other identical, high-use data. KSM allows for greater guest density of identical or similar guest operating systems by avoiding memory duplication. The concept of shared memory is common in modern operating systems. For example, when a program is first started, it shares all of its memory with the parent program. When either the child or parent program tries to modify this memory, the kernel allocates a new memory region, copies the original contents and allows the program to modify this new region. This is known as copy on write. KSM is a Linux feature which uses this concept in reverse. KSM enables the kernel to examine two or more already running programs and compare their memory. If any memory regions or pages are identical, KSM reduces multiple identical memory pages to a single page. This page is then marked copy on write. If the contents of the page is modified by a guest virtual machine, a new page is created for that guest. This is useful for virtualization with KVM. When a guest virtual machine is started, it only inherits the memory from the host qemu-kvm process. Once the guest is running, the contents of the guest operating system image can be shared when guests are running the same operating system or applications. KSM allows KVM to request that these identical guest memory regions be shared. KSM provides enhanced memory speed and utilization. With KSM, common process data is stored in cache or in main memory. This reduces cache misses for the KVM guests, which can improve performance for some applications and operating systems. Secondly, sharing memory reduces the overall memory usage of guests, which allows for higher densities and greater utilization of resources. NOTE In Red Hat Enterprise Linux 6.7 and later, KSM is NUMA aware. This allows it to take NUMA locality into account while coalescing pages, thus preventing performance drops related to pages being moved to a remote node. Red Hat recommends avoiding crossnode memory merging when KSM is in use. If KSM is in use, change the /sys/kernel/mm/ksm/merge_across_nodes tunable to 0 to avoid merging pages across NUMA nodes. This can be done with the virsh node-memory-tune --shmmerge-across-nodes 0 command. Kernel memory accounting statistics can eventually contradict each other after large amounts of cross-node merging. As such, numad can become confused after the KSM daemon merges large amounts of memory. If your system has a large amount of free memory, you may achieve higher performance by turning off and disabling the KSM daemon. Refer to Chapter 9, NUMA" for more information on NUMA. 30

35 CHAPTER 8. MEMORY IMPORTANT Ensure the swap size is sufficient for the committed RAM even without taking KSM into account. KSM reduces the RAM usage of identical or similar guests. Overcommitting guests with KSM without sufficient swap space may be possible, but is not recommended because guest virtual machine memory use can result in pages becoming unshared. Red Hat Enterprise Linux uses two separate methods for controlling KSM: The ksm service starts and stops the KSM kernel thread. The ksmtuned service controls and tunes the ksm service, dynamically managing same-page merging. ksmtuned starts the ksm service and stops the ksm service if memory sharing is not necessary. When new guests are created or destroyed, ksmtuned must be instructed with the retune parameter to run. Both of these services are controlled with the standard service management tools. NOTE KSM is off by default on Red Hat Enterprise Linux The KSM Service The ksm service is included in the qemu-kvm package. When the ksm service is not started, Kernel same-page merging (KSM) shares only 2000 pages. This default value provides limited memory-saving benefits. When the ksm service is started, KSM will share up to half of the host system's main memory. Start the ksm service to enable KSM to share more memory. # systemctl start ksm Starting ksm: [ OK ] The ksm service can be added to the default startup sequence. Make the ksm service persistent with the systemctl command. # systemctl enable ksm The KSM Tuning Service The ksmtuned service fine-tunes the kernel same-page merging (KSM) configuration by looping and adjusting ksm. In addition, the ksmtuned service is notified by libvirt when a guest virtual machine is created or destroyed. The ksmtuned service has no options. # systemctl start ksmtuned Starting ksmtuned: [ OK ] The ksmtuned service can be tuned with the retune parameter, which instructs ksmtuned to run tuning functions manually. 31

36 Virtualization Tuning and Optimization Guide The /etc/ksmtuned.conf file is the configuration file for the ksmtuned service. The file output below is the default ksmtuned.conf file: # Configuration file for ksmtuned. # How long ksmtuned should sleep between tuning adjustments # KSM_MONITOR_INTERVAL=60 # Millisecond sleep between ksm scans for 16Gb server. # Smaller servers sleep more, bigger sleep less. # KSM_SLEEP_MSEC=10 # KSM_NPAGES_BOOST - is added to the `npages` value, when `free memory` is less than `thres`. # KSM_NPAGES_BOOST=300 # KSM_NPAGES_DECAY - is the value given is subtracted to the `npages` value, when `free memory` is greater than `thres`. # KSM_NPAGES_DECAY=-50 # KSM_NPAGES_MIN - is the lower limit for the `npages` value. # KSM_NPAGES_MIN=64 # KSM_NPAGES_MAX - is the upper limit for the `npages` value. # KSM_NPAGES_MAX=1250 # KSM_THRES_COEF - is the RAM percentage to be calculated in parameter `thres`. # KSM_THRES_COEF=20 # KSM_THRES_CONST - If this is a low memory system, and the `thres` value is less than `KSM_THRES_CONST`, then reset `thres` value to `KSM_THRES_CONST` value. # KSM_THRES_CONST=2048 # uncomment the following to enable ksmtuned debug information # LOGFILE=/var/log/ksmtuned # DEBUG=1 Within the /etc/ksmtuned.conf file, npages sets how many pages ksm will scan before the ksmd daemon becomes inactive. This value will also be set in the /sys/kernel/mm/ksm/pages_to_scan file. The KSM_THRES_CONST value represents the amount of available memory used as a threshold to activate ksm. ksmd is activated if either of the following occurs: The amount of free memory drops below the threshold, set in KSM_THRES_CONST. The amount of committed memory plus the threshold, KSM_THRES_CONST, exceeds the total amount of memory KSM Variables and Monitoring Kernel same-page merging (KSM) stores monitoring data in the /sys/kernel/mm/ksm/ directory. Files in this directory are updated by the kernel and are an accurate record of KSM usage and statistics. 32

37 CHAPTER 8. MEMORY The variables in the list below are also configurable variables in the /etc/ksmtuned.conf file, as noted above. Files in /sys/kernel/mm/ksm/: full_scans Full scans run. merge_across_nodes Whether pages from different NUMA nodes can be merged. pages_shared Total pages shared. pages_sharing Pages currently shared. pages_to_scan Pages not scanned. pages_unshared Pages no longer shared. pages_volatile Number of volatile pages. run Whether the KSM process is running. sleep_millisecs Sleep milliseconds. These variables can be manually tuned using the virsh node-memory-tune command. For example, the following specifies the number of pages to scan before the shared memory service goes to sleep: # virsh node-memory-tune --shm-pages-to-scan number KSM tuning activity is stored in the /var/log/ksmtuned log file if the DEBUG=1 line is added to the /etc/ksmtuned.conf file. The log file location can be changed with the LOGFILE parameter. Changing the log file location is not advised and may require special configuration of SELinux settings Deactivating KSM Kernel same-page merging (KSM) has a performance overhead which may be too large for certain environments or host systems. KSM may also introduce side channels that could be potentially used to leak information across guests. If this is a concern, KSM can be disabled on per-guest basis. 33

38 Virtualization Tuning and Optimization Guide KSM can be deactivated by stopping the ksmtuned and the ksm services. However, this action does not persist after restarting. To deactivate KSM, run the following in a terminal as root: # systemctl stop ksmtuned Stopping ksmtuned: [ OK ] # systemctl stop ksm Stopping ksm: [ OK ] Stopping the ksmtuned and the ksm deactivates KSM, but this action does not persist after restarting. Persistently deactivate KSM with the systemctl commands: # systemctl disable ksm # systemctl disable ksmtuned When KSM is disabled, any memory pages that were shared prior to deactivating KSM are still shared. To delete all of the PageKSM in the system, use the following command: # echo 2 >/sys/kernel/mm/ksm/run After this is performed, the khugepaged daemon can rebuild transparent hugepages on the KVM guest physical memory. Using # echo 0 >/sys/kernel/mm/ksm/run stops KSM, but does not unshare all the previously created KSM pages (this is the same as the # systemctl stop ksmtuned command). 34